A decade ago, a disintegrin and metalloproteinase 10 (ADAM10) was identified as an α-secretase and as a key proteinase in the processing of the amyloid precursor protein. Accordingly, the important role that it plays in Alzheimer’s disease was manifested. Animal models with an overexpression of ADAM10 revealed a beneficial profile of the metalloproteinase with respect to learning and memory, plaque load and synaptogenesis. Therefore, ADAM10 presents a worthwhile target with respect to the treatment of a neurodegenerative disease such as Morbus Alzheimer. Initially, ADAM10 was suggested to be an enzyme, shaping the extracellular matrix by cleavage of collagen type IV, or to be a tumour necrosis factor α convertase. In a relatively short time, a wide variety of additional substrates (with amyloid precursor protein probably being the most prominent) has been identified and the search is still ongoing. Hence, any side effects concerning the therapeutic enhancement of ADAM10 α-secretase activity have to be considered. The present review summarizes our knowledge about the structure and function of ADAM10 and highlights the opportunities for enhancing the expression and/or activity of the α-secretase as a therapeutic target.
Identification of ADAM10 as a functional α-secretase
A disintegrin and metalloproteinase 10 (ADAM10) originally came into focus in genetical and biochemical research as a peptide sequence purified from bovine brain myelin membrane preparations , and was referred to as MADM (i.e. mammalian disintegrin-metalloprotease). Accidentally, this metalloproteinase was identified via an artifact resulting from in vitro studies: it has been described as a proteinase for the cytosolic myelin basic protein , which is a rather unphysiological substrate for the type I transmembrane enzyme ADAM10. Further studies revealed that ADAM10 is expressed in a wide variety of tissues either in Bos taurus  and, more interestingly, in distinct areas of the human brain [4,5] and peripheral structures [6,7]. Striking similarity concerning the inhibitory profile of ADAM10  with the putative α-secretase  suggested a more physiological role for its enzymatic activity: overexpression of the ADAM10 cDNA in HEK293 cells first identified its function as an amyloid precursor protein (APP) cleaving α-secretase , which subsequently was verified in vivo. Alzheimer’s disease (AD) model mice, which were crossbred with ADAM10 transgenic mice, revealed a strongly attenuated plaque pathology and an enhanced production of the α-secretase derived soluble cleavage product APPs-α . Furthermore, these mice had an increased learning and memory potential , which might correlate with the observed enhanced cholinergic and glutamatergic synaptogenesis . By contrast, mice with a dominant negative mutant of ADAM10 had lowered amounts of APPs-α, accompanied by an enhanced amount of plaques  and learning deficiencies in the Morris water maze test . In summary, what began with a fallacious observation ended up with the discovery of an enzyme that might have implications for a therapeutic approach in AD [13–15].
Protein structure and gene organization of ADAM10
The enzyme ADAM10 belongs to the subgroup of metzincins within the zinc proteinases family. The typical multidomain structure of ADAM10 as a type I integral transmembrane protein consists of a prodomain, a catalytical domain with a conserved zinc binding sequence, a cysteine-rich disintegrin-like domain, a transmembrane domain and a rather short cytoplasmic domain (Fig. 1).
The nascent protein itself is not functional and is produced as a zymogene. After cleavage of the signalling sequence, ADAM10 enters the secretory pathway to be processed and thereby activated by the proprotein convertases furin or PC7 . This constitutive processing has been demonstrated for the prodomains of several ADAMs [17–19]. Regarding ADAM10, the prodomain was revealed to exhibit a dual function: the separately expressed prodomain was capable of inactivating endogenous ADAM10 in cell culture experiments but overexpressed ADAM10 without its prodomain was inactive . By contrast, coexpression of the prodomain in trans rescued the activity of the deletion mutant of ADAM10 without the intramolecular prodomain . In addition, the recombinant murine prodomain purified from Escherichia coli acts as a potent and selective competitive inhibitor in experiments performed in vitro . This implicates that the prodomain of ADAM10 acts not only as a transient inhibitor, but also as an internal chaperone in the maturation of the enzyme. Accordingly, the viral delivery of furin into the brain of AD model mice increased α-secretase activity and reduced β-amyloid protein (Aβ) production in infected brain regions , demonstrating the in vivo relevance of the removal of the prodomain of ADAM10. Recently, by reciprocal coimmunoprecipitation, tetraspanin 12 was identified as an interaction partner for ADAM10 that enhances α-secretase shedding of APP, probably by regulating maturation of the prodomain of ADAM10 .
The catalytical domain of ADAM10 contains a typical zinc-binding consensus motif (HEXGHXXGXXHD; Fig. 1) and the point mutation E384A, which compromises this motif, leads to a substantial decrease in APPs-α secretion in HEK cells and in mice [10,23]. Glycosylation sites in the catalytic and disintegrin domain contain high-mannose as well as complex-type N-glycans, and a mutation at the N-glycosylation site N439 increased ADAM10s susceptibility to proteolytical degradation .
Although the removal of the disintegrin domain of ADAM10 did not grossly affect shedding of APP in cell culture experiments , cleavage of some substrate molecules is likely to be influenced by noncatalytical domains. For example, epidermal growth factor cleavage is at least partially impaired in ADAM10−/− cells overexpressing a cytoplasmic domain deletion mutant of ADAM10 . In accordance with this finding, the cytoplasmic domain of ADAM10 contains an IQ consensus binding site for calmodulin that afflicts maturation of the proteinase . Additionally, ADAM10 has been shown to be activated by a calcium ionophore and the calmodulin inhibitor trifluoroperazine [26,27]. The cytoplasmic domain of ADAM10 furthermore contains two proline-rich putative Src homology 3 (SH3) binding domains, from which the juxtamembrane domain affects basolateral localization of ADAM10 in epithelial cells . In neurones, the SH3 binding domains direct ADAM10 via binding to synapse-associated protein-97 to the postsynaptic membrane .
In 1997, the gene locus for ADAM10 was matched to chromosome 15 in humans (15q21.3-q23) and chromosome 9 in mice [30,31]. Subsequently, it took 8 years to achieve further gene structure analysis and potential identification of transcription factor binding sites . We now know that the human, mouse and rat genes, which comprise ∼ 160 kb, include a highly homologous sequence within the first 500 bp upstream of either translation initiation site. Deletion analysis defined nucleotides −508 to −300 bp as the human core promoter. This promoter was also identified as a TATA-less promoter with functional binding sites for Sp1, USF and retinoic acid receptors [32,33]. The functional promoter of ∼ 2 kb displayed activity in various human cell lines, such as HEK293, HepG2 or SH-SY5Y, which reflects the ubiquitous basal expression of the endogenous ADAM10.
Single nucleotide polymorphism analyses of the promoter region of 104 AD patients versus control patients (n = 84) did not lead to significant statistical differences . In addition, an independent recent study, genotyping 27 single nucleotide polymorphisms covering the entire gene for ADAM10 in a larger cohort of patients (n = 438 AD; n = 290 control), revealed no single-marker or haplotypic association with the disease . This indicated that the gene for ADAM10 probably does not constitute a major risk with regard to AD. Nevertheless, a very recent study of 1439 DNAs from 436 multiplex AD families yielded significant evidence for an association of AD with the metalloproteinase with respect to two mutations: Q170H and R181G . Both mutations are located close to the cysteine switch within the prodomain and the proprotein convertase recognition site (Fig. 1), which explains their strong impact on enzyme functionality: Chinese hamster ovary cells stably overexpressing mutated ADAM10 showed strongly attenuated α-secretase activity . Although both mutations are rare (segregation in seven AD families out of 1004) and are only partially penetrant, these results give support to the hypothesis that the human gene for ADAM10 plays a role in the aetiology of AD.
ADAM10 and tumour necrosis factor α(TACE): the ill-matched couple
Three members of the ADAM family have been shown to act as α-secretase [8,36,37]: ADAM9, ADAM10 and ADAM17 (TACE). Overexpression of ADAM9 has been reported to increase the basal and protein kinase C (PKC) dependent APPs-α release , although the purified enzyme failed to cleave a synthetic peptide at the major α-secretase cleavage-site . Additionally, mice lacking ADAM9 revealed no differences in the production of the α-secretase cleavage product of APP . The impact of ADAM9 promoter polymorphism on sporadic AD, which has been described recently , might therefore rely on a more indirect mechanism: ADAM9 has been shown to proteolytically process ADAM10 [40–42]. By contrast to ADAM9, ADAM10 was found to have constitutive and regulated α-secretase activity as well as many other properties expected for the α-secretase [8,10]. Moreover, in situ hybridization analysis in human cortical neurones provided evidence for the coexpression of APP with ADAM10, suggesting that this proteinase is most likely the physiologically relevant α-secretase . Finally, experiments performed with ADAM17 (TACE)-deficient cells indicated a participation of TACE in the regulated, PKC-stimulated [37,43] and the constitutive α-secretase pathway [44,45]. To our knowledge, there are no published reports about TACE acting as an in vivo APP-sheddase in transgenic mice, although TACE-positive neurones are found to colocalize with amyloid plaques in AD brains supporting its role as an α-secretase .
On the basis of these results, it can be concluded that ADAM10 and TACE are the major sheddases that balance the β-site amyloid precursor protein cleaving enzyme-driven generation of Aβ peptides. This is consistent with the close structural relationship of both metalloproteinases: although TACE of human origin has ∼ 30% amino acid identity relative to bovine ADAM10, it only shows ∼ 15% identity with ADAM9 . Additionally, only those two ADAMs lack the RX(6)DLPEFα(9)β(1) integrin binding motif, which is contained in the other members of the proteinase family . Nevertheless, there are significant differences between ADAM10 and TACE that probably allow a specific modulation of one of them for therapeutic approaches. TACE not only differs in the consensus sequence of its disintegrin domain from ADAM10 or by including a Crambin-like domain , but also in its regulation. Several studies have described the treatment of cellular cultures with a distinct outcome for either TACE or ADAM10 activity: for example, incubation with phorbol 12-myristate 13-acetate increased the turnover of TACE in Jurkat cells  and diminished the amount of mature TACE in HEK293 as well as in SH-SY5Y cells . Interestingly, these cell lines did not show altered amounts of ADAM10, suggesting a significant difference in the cellular stability of the mature enzyme forms after treatment with 4β-phorbol 12-myristate 13-acetate . In addition, ADAM10 and TACE vary in their reaction to cellular differentiation by retinoic acid [50,51] and active site determinants of substrate recognition .
ADAM10: not particular about its substrates?
For the enzyme ADAM10, more than 40 substrates have been identified that belong to three different classes of membrane bound proteins . Most of them are type I transmembrane proteins such as APP , APP-like protein 2  or the receptor for glycosylation end products [54,55]. Type II transmembrane proteins such as the apoptosis-inducing Fas ligand [56,57] or Bri2  have also been reported to be shed by ADAM10. Additionally, at least three glycosylphosphatidylinositol (GPI)-anchored proteins are candidate substrates for ADAM10: the metastasis-associated protein C4.4A was characterized by a proteome technique as a substrate of ADAM10 . Furthermore, the GPI-anchored neuronal guidance molecule ephrin A5 is cleaved by ADAM10 upon binding to its receptor EphA3, leading to termination of the receptor–ligand interaction . Third, from cell culture experiments, the prion protein PrPc was suggested to be processed by ADAM10 [40,61] and the abundance of the PrP cleavage product C1 was associated with mature ADAM10 within a small set of human cerebral cortex samples . However, in vivo overexpression of ADAM10 in mice reduced all cellular prion protein species instead of generating enhanced amounts of cleavage products .
The substrates of ADAM10 show a wide range of cellular function
ADAM10 cleaves proteins that affect cell migration (N-cadherin ; transmembrane chemokines ) and cell proliferation (CXCL16 [66–68]). It also sheds proteins with functions in either the immune system (low affinity immunoglobulin E receptor [69,70]; vascular endothelial cadherin ) or in cell signalling (Delta ; Notch ). Most effects, provoked by ADAM10 shedding activity, have been associated with the huge N-terminal ectodomains of the substrates of ADAM10 that are released into the intercellular fluid upon cleavage. However, some effects have clearly been matched to the intracellular domains of the substrates: ectodomain shedding by ADAM10 is followed by regulated intramembrane proteolysis. After cleavage of the Notch receptor by ADAM10, γ-secretase releases a small intracellular part of Notch, which then translocates to the nucleus and acts as a transcription factor [74–76]. With regard to Bri2, the ADAM10-derived cleavage is followed by signal peptide peptidase-like protease activity, also resulting in the release of a small Bri2 fragment into the cell body .
In summary, ADAM10 has a repertoire of different protein substrates hampering the development of therapeutic strategies that target specifically APP by ADAM10. However, not all substrates described as being cleaved in the in vitro system have been confirmed in vivo. Mutagenesis experiments have depicted at least three residues in the S1′ pocket of ADAM10 that strongly influence substrate specificity and also limit the number of substrates . Additional interactions of ADAM10 noncatalytical domains with the substrate or with adaptor molecules, as previously described for the recognition of ephrins , also appear to be important for targeting ADAM10 to a distinct substrate in the physiological context.
Regulators of ADAM10 expression and catalytical activity
Because of the above-mentioned involvement of ADAM10 in a wide range of cellular functions, it is obvious to consider its therapeutic potential in various diseases such as cancer or AD. ADAM10 has been shown to cleave tumour-associated substrates such as MICA  or C4.4A  and to be linked to progression of certain cancer types such as prostate or breast cancer [78–80]. Furthermore, it plays a role in metastasis of human colon cancer cells . Therefore, the inhibition of ADAM10 might be helpful in cancer treatment in certain contexts . By contrast, ADAM10 overexpression or activation in the brain might be beneficial for the treatment of neurodegenerative diseases, in particular AD: this progressive disorder of the brain goes ahead with the loss of synaptic junctions and neuronal cells. For ADAM10 overexpressing mice, it has been demonstrated that cortical synaptogenesis is enhanced , long-term potentiation deficiency in AD model mice is rescued  and learning, as well as memory, is positively influenced by ADAM10 . Studies with the dominant negative form of ADAM10 in a mouse model of AD revealed that the enzymatic activity of ADAM10 is required to counteract cognitive deficits . In addition, axonal guidance is conveyed by the metalloproteinase, as has been shown for retinal and peripheral axons [84,85], and ADAM10 regulates axon withdrawal by ephrin cleavage [60,86].
It remains a matter of controversy as to whether there is a substantial decline of neuronal ADAM10 in ageing or in the pathological context: healthy, ageing human fibroblasts did not reveal lowered amounts of ADAM10 during senescence , although its specific cleavage product APPs-α was decreased. Another study demonstrated ADAM10 mRNA to be upregulated in cases of presenile dementia but to be downregulated in the brain of AD patients . A decrease for ADAM10 and APPs-α was confirmed in human platelets [88,89] as well as for APPs-α in the cerebrospinal fluid of AD patients. Additionally, a recent study revealed that colocalization of ADAM10 and one of its potential regulators (i.e. nardilysin) is reduced in AD compared to healthy aged brains .
With regard to these reports and to studies with ADAM10 overexpression in a mouse AD model , in principal, the enhancement of ADAM10 activity and/or amount in the patient’s brain appears to be valuable. How can this be achieved? Different approaches appear to be promising, such as interfering with the transcription/translation of ADAM10 or regulating its enzymatic capacity by influencing the membrane physiology or via protein interactions (Fig. 2).
A first point of intervention within the biosynthetic pathway of ADAM10 is provided by directly interfering with the expression of the gene for ADAM10: the promoter region of the gene for ADAM10 has been characterized in detail  and in silico analyses have provided a multitude of transcription factor binding sites. One of the putative binding sites for retinoic acid receptors located at −302 and −203 bp has been demonstrated to be functional by electrophoretic mobility shift assay, promoter assays and APPs-α secretion in human neuronal cells [32,50]. In addition, acitretin, which is an accredited synthetic retinoid drug, lowered Aβ peptide generation in AD model mice and enhanced APPs-α secretion . Acitretin, which is already used in the long-term treatment of patients suffering from skin diseases withdraws all-trans retinoic acid from its cellular retinoic acid binding protein and makes it available for activating the corresponding nuclear receptors. In the case of ADAM10 regulation, cell culture studies with a variety of ligands for nuclear receptors narrowed the receptors involved down to a nonpermissive retinoic acid receptor–retinoid X receptor heterodimer .
Another approach is offered by targeting the nascent ADAM10 molecules during maturation within the cell. Enhancement of the expression of a proprotein convertase such as furin will increase ADAM10 maturation and α-secretase activity . A further cleavage of ADAM10 has been described in close proximity to and within its transmembrane domain [40–42]. This is a result of metalloproteinases ADAM9 and 15 acting on ADAM10 to release a soluble sADAM10 from the cell surface. sADAM10 was incapable of shedding cell-associated amyloid precursor protein , whereas it cleaved a synthetic peptide substrate [41,42] and endogenous prion protein in cell culture experiments . Because it is still unclear whether soluble ADAM10 and the transmembrane variant cleave the same substrates and whether they have the same catalytic properties in vivo, this type of regulation has to be elucidated further. Acetylcholine esterase inhibitors, which are already used in the symptomatic treatment of AD, enhance the transport of ADAM10 to the cell surface and the non-amylodogenic cleavage of APP [91,92].
ADAM10 has also been shown to be regulated by the lipid composition of the plasma membrane. While cholesterol depletion enhanced its activity [93,94], targeting ADAM10 via an artificial GPI-anchor to cholesterol-rich domains inhibited its enzymatic function . In human cells, the amount and activity of ADAM10 was enhanced by statin application . However, the outcomes of clinical trials with the cholesterol lowering statins are not unambiguous: several studies have reported a protective effect of statins against AD [96,97], although this could not be confirmed in others [98,99]. Nevertheless, in the prospective, population-based Rotterdam study comprising ∼ 7000 participants, the use of statins was associated with a lower risk of AD , preserving the hope of a therapeutic value for statins in AD therapy. Further evidence for lipids acting as modulators of α-secretase activity is provided by a study demonstrating that type III secretory phospholipase A and arachidonic acid increased APPs-α production most likely by enhancing substrate availability at the cell surface .
Another approach to activate ADAM10 could rest on noncovalent protein interaction partners of ADAM10. The tissue inhibitors of metalloproteinases 1 and 3 have been shown to inhibit ADAM10 in vitro  and the reversion-inducing cysteine-rich protein with Kazal motifs also comprises a physiological ADAM10 inhibitor . By contrast, for the N-arginine dibasic convertase (nardilysin), an activating property for ADAM10-mediated APP α-secretase cleavage and tumour necrosis factor α cleavage has been reported [104,105]. The same holds true for the tetraspanins: tetraspanin 12 increases maturation and activity of ADAM10  and ADAM10 has been suggested as a component of the ‘tetraspanin web’ , which also scaffolds heterotrimeric G protein-coupled receptors (GPCRs) . For the development of drugs interacting with those proteins and thereby modulating ADAM10 activity, further studies are necessary.
An appropriate strategy for targeting ADAM10 is presented by directly stimulating the ADAM10 activity by ligands of GPCRs. For example, the GPCR ligands LPA and bombesin induced ADAM10-driven epidermal growth factor receptor transactivation  and shedding of the thyrotropin receptor by ADAM10 was mediated by its ligand thyrotropin . At least in cell culture, the α-secretase cleavage of APP is inducible by the neuropeptide pituitary adenylate cyclase-activating peptide (PACAP), which involves signalling via mitogen-activated protein kinase and phosphatidylinositol 3-kinase . These results are of special interest because the neuropeptide PACAP offers the opportunity of locally activating the PAC1 receptor and α-secretase in the brain. This also holds true for the serotonin 5-hydroxytryptamine (5-HT4) receptor, which increases memory and learning: the 5-HT4(ε) receptor isoform induced α-secretase activity by the cAMP-regulated guanine exchange factor Epac and the small GTPase Rac [111,112]. This recently led to synthesis and evaluation of novel 5-HT4-agonists; two of them increased APPs-α production in the cortex and hippocampus of mice and exhibited neuroprotective properties .
Therefore, GPCR ligands offer an interesting opportunity in regulating ADAM10, even if the signalling pathways have not yet been elucidated in every detail. Another signalling pathway regulating ADAM10 activity is connected with the PKC: in various in vitro studies, it has been demonstrated that PKC or certain isoforms of PKC stimulate the α-secretase [114–116] (for the role of PKC in AD, see ) and this has been confirmed in AD model mice (e.g. bryostatin 1) .
ADAM10 as target for AD therapy: lessons learned from transgenic mice
In summary, several independent strategies for enhancing the amount or the catalytic activity of ADAM10 have been performed or are conceivable. The crucial question remaining is whether there are side effects connected with enhanced ADAM10 activity in the brain or in peripheral structures. ADAM10 mono-transgenic mice with a permanent neuronal overexpression of ADAM10 to various extent were inconspicuous in morphology, breeding and in daily handling . This indicates that, by overexpression of ADAM10 in the brain, the homeostasis of the entire organism is not grossly affected. A more detailed behavioural examination showed that ADAM10 moderately overexpressing mice performed similar to controls with respect to basal activity, exploration and anxiety. In the Morris water maze hidden platform task, however, ADAM10 mono-transgenic mice showed thigmotaxis with floating behaviour, indicating differences in motivation . Therefore, with respect to learning and memory, mono-transgenic ADAM10 mice displayed no specific phenotype. By contrast, overexpression of ADAM10 in an AD mouse model with mutated human APP created bi-transgenic mice with a clear improvement of memory and alleviation of learning deficits .
A recent microarray study  revealed that there was only a moderate alteration of gene expression in moderately ADAM10 overexpressing adult mice. Genes coding for pro-inflammatory or pro-apoptotic proteins were not over-represented among differentially regulated genes and, indeed, a decrease of inflammation markers was observed. ADAM10 participates also in the activation of Notch1 signalling by cleaving the extracellular portion of this receptor upon ligand binding. Young ADAM10 transgenic mice at postnatal day 15 showed a 40% induction of expression of the gene for Hes5, whereas a 50% reduction in mice overexpressing the dominant negative variant of the enzyme was reported . Nevertheless, in adult mice, no significant effects with respect to the amount of Notch1 target gene Hes5 mRNA were obtained, suggesting an attenuation of the signalling cascade during ageing. Because ADAM10-based AD therapy will take place in elderly people, interference with this important developmental signalling pathway does not appear to hamper such an approach.
Regarding prion diseases, upregulation of ADAM10 might also be beneficial: the reduction of all species of the prion protein in ADAM10 overexpressing mice was accompanied by a prolonged survival time of the mice after Scrapie infection . In addition, Akt phosphorylation as a marker for survival signals in neuronal cells  was not affected in ADAM10 moderately overexpressing mice . Furthermore, the thickness of the myelin sheath was not altered by ADAM10 overexpression, demonstrating that neuregulin-1 acting as a modulator of this developmental event is not a substrate of ADAM10 . In mice characterized by high levels of overexpressed ADAM10, however, phosphorylation of Akt was reduced to ∼ 50% compared to wild-type mice and tomacula-like structures (i.e. local myelin thickenings) were observed . In addition, mice with high ADAM10 overexpression showed more seizures and stronger neuronal damage and inflammation than wild-type mice upon kainate treatment . By contrast, in the presence of its substrate APP in doses exceeding the endogenous level, ADAM10 revealed a protective effect .
If we consider all of the results obtained concerning increased ADAM10 activity in vivo, it can be concluded that this approach might be a valuable alternative to other strategies, such as the inhibition of β- or γ-secretase or immunization, for the treatment of AD. However, α-secretase activation must be moderate and closely monitored.
The authors’ own work is supported by the Federal Ministry of Education and Research (BMBF) in the Framework of the National Genome Research Network (NGFN), Förderkennzeichen FKZ01GS08130.